Why should I trust you? Influence of explanation design on consumer behavior in AI-based services

Florence Nizette, Wafa Hammedi, Allard van Riel, Nadia Steils

Track: Type B (Encore Abstracts)

Keywords: Design, Recommendations, Credence service, AI, XAI, Explanations

Abstract:

Purpose: This study investigates how the format of explanations in artificial intelligence (AI)-based services shapes consumer perceptions and behaviors. Specifically, it examines the effects of explanation detail (low vs. high) and consumer control (automatic vs. on-demand) on perceived understanding, assurance, trust, acceptance and purchase intention. By doing so, the study aims to provide actionable insights for service providers seeking to optimize explanatory strategies that enhance consumer evaluations of AI-based services in high-credence contexts.

Design/methodology/approach: Building on research in explainable AI (XAI) and information overload, a conceptual model was developed (Bauer et al., 2023; Rai, 2020). Two between-subjects experiments were conducted, manipulating explanation detail and consumer control in the context of AI-based services. Data was analyzed using partial least squares (PLS) regression to assess direct, moderating and mediating effects.

Findings: The results reveal that detailed explanations significantly improve perceived understanding and assurance, which in turn enhance consumer trust, acceptance and purchase intention. Importantly, consumer control over explanations negatively moderates the relationship between detail and understanding: when consumers are given high control, the benefits of detailed explanations diminish, likely due to perceived information load. The analyses further show that perceived assurance emerges as a key antecedent of trust, enhancing perceptions of AI competence and integrity.

Theoretical contributions: This study makes four contributions. First, it advances a consumer-centric perspective on XAI, highlighting explanations as a mechanism to reduce resistance to AI in high-credence services (Chen et al., 2024; Gaczek, et al., 2023). Second, it extends the literature on information overload by showing that detailed explanations can increase trust and acceptance when consumer control is limited, challenging assumptions that detail necessarily overwhelms consumers (He et al., 2023). Third, it identifies perceived assurance as a critical mediating mechanism linking explanations to trust, thus extending prior work on XAI and trust formation (Haque et al., 2023). Finally, it offers a nuanced account of how the dimensions of trust (competence, integrity and human centricity) are shaped by explanation design.

Practical implications: For managers, the findings highlight the importance of tailoring explanatory strategies by balancing detail with control. Service providers should prioritize clear, detailed explanations that reveal the rationale behind AI services while carefully managing the degree of consumer control.

Originality/value: This research contributes to the emerging discourse on consumer-centric XAI by elucidating the interplay between explanation detail and control, particularly in high-credence services where consumers rely heavily on provider expertise. By demonstrating the important role of perceived assurance in trust formation, the study offers new insights into designing explanations that inform and also reassure consumers, thereby increasing their acceptance of AI-based services (Longoni et al., 2019; Kim et al., 2024).

References available upon request

Serve As Reviewer: ~Florence_Nizette1

Submission Number: 41